Machine learning algorithms have been successfully applied to learning classifiers in many domains such as computer vision, fraud detection, and brain image analysis. Typically, classifiers are trained to predict a class value given a set of labeled training data that includes all possible class values, and sometimes additional unlabeled training data. Little research has been performed where the possible values for the class variable include values that have been omitted from the training examples. This is an important problem setting, especially in domains where the class value can take on many values, and the cost of obtaining labeled examples for all values is high. We show that the key to addressing this problem is not predicting the held-out classes directly, but rather by recognizing the semantic properties of the classes such as their physical or functional attributes. We formalize this method as zero-shot learning and show that by utilizing semantic knowledge mined from large text corpora and crowd-sourced humans, we can discriminate classes without explicitly collecting examples of those classes for a training set. As a case study, we consider this problem in the context of thought recognition, where the goal is to classify the pattern of brain activity observed from a non-invasive neural recording device. Specifically, we train classifiers to predict a specific concrete noun that a person is thinking about based on an observed image of that person’s neural activity. We show that by predicting the semantic properties of the nouns such as “is it heavy?” and “is it edible?”, we can discriminate concrete nouns that people are thinking about, even without explicitly collecting examples of those nouns for a training set. Further, this allows discrimination of certain nouns that are within the same category with significantly higher accuracies than previous work. In addition to being an important step forward for neural imaging and braincomputer- interfaces, we show that the zero-shot learning model has important implications for the broader machine learning community by providing a means for learning algorithms to extrapolate beyond their explicit training set.

Most of the work investigating the representation of concrete nouns in the brain has focused on the locations that code the information. We present a model to study the contributions of perceptual and semantic features to the neural code representing concepts over time and space. The model is evaluated using magnetoencephalography data from different paradigms and not only corroborates previous findings regarding a distributed code, but provides further details about how the encoding of different subcomponents varies in the space-time spectrum. The model also successfully generalizes to novel concepts that it has never seen during training, which argues for the combination of specific properties in forming the meaning of concrete nouns in the brain. The results across paradigms are in agreement when the main differences among the experiments (namely, the number of repetitions of the stimulus, the task the subjects performed, and the type of stimulus provided) were taken into consideration. More specifically, these results suggest that features specific to the physical properties of the stimuli, such as word length and right-diagonalness, are encoded in posterior regions of the brain in the first hundreds of milliseconds after stimulus onset. Then, properties inherent to the nouns, such as is it alive? and can you pick it up?, are represented in the signal starting at about 250 ms, focusing on more anterior parts of the cortex. The code for these different features was found to be distributed over time and space, and it was common for several regions to simultaneously code for a particular property. Moreover, most anterior regions were found to code for multiple features, and a complex temporal profile could be observed for the majority of properties. For example, some features inherent to the nouns were encoded earlier than others, and the extent of time in which these properties could be decoded varied greatly among them. These findings complement much of the work previously described in the literature, and offer new insights about the temporal aspects of the neural encoding of concrete nouns. This model provides a spatiotemporal signature of the representation of objects in the brain. Paired with data from carefully-designed paradigms, the model is an important tool with which to analyze the commonalities of the neural code across stimulus modalities and tasks performed by the subjects.

Three collimators were designed and built with mouse irradiation in mind. They were made to have a shapeable aperture that could deliver strange or complex dose shapes to spots on the animals. Current collimators are either custom-built, expensive, or only provide limited dose shapes. These provide self-customizable collimation at a minimum of cost. A mouse holder was also devised to reproducibly hold a mouse under the collimator. Construction of the holder and of each collimator is also covered. Each collimator was analyzed for flaws, holes, and penumbral width across various shapes of interest. A Norelco MG300 X-ray generator at the Texas A&M Nuclear Science Center was used in these tests. The lead collimator had a radial penumbra of 1.2 mm. The clay / lead shot collimator had a radial penumbra of 1.6 mm. The brass collimator had a radial penumbra of 1.75 mm. Vertical and horizontal penumbras are dependent on distance from the center of the beam. All readings are + 0.3 mm according to the resolution of the scanner used in this experiment. Each collimator is useful for different purposes.

Standardization, flexibility and transparency had been the eminent keywords of Modern Architecture that have also provided material for the re-interpretation and re-production of several discussions on concepts like &lsquo / form&rsquo / and &lsquo / function&rsquo / in architecture. This study is an inquiry into a number of different interpretations that scrutinize the intricate relationship between architectural form and function in the 20th century with respect to the concepts defined above. The conceptualization of this thesis will be based on the assumption that function is an inadequate and weak concept for the generation of architectural form. It is in this context that form is accepted as an architectural &ldquo / fragment&rdquo / that attains its autonomy not from the strictures imposed by other architectural fragment like &lsquo / function&rsquo / , &lsquo / structure&rsquo / , &lsquo / program&rsquo / and &lsquo / site / but from an infinite number of internal relationships or &lsquo / design tools&rsquo / as we call. The aim of this study is to propose a project for &ldquo / A Multi &ndash / Functional Hall at METU&rdquo / which has the capability of expanding its limits in future, both physically and conceptually / and investigate the &lsquo / design tools&rsquo / that will direct the steps of transformations in the process of architectural production. Within this framework, the investigation is concerned with the production of a flexible, unstable and indeterminant building, focused on the concepts of &lsquo / transformation of space&rsquo / and &lsquo / constant change&rsquo / , that could be re-designed and re-generated in respect to possible future transformations in the program. Thus, architectural production is defined merely as a &lsquo / step&rsquo / or a &lsquo / snap-shot&rsquo / , controlled by the design tools suggesting the solutions for an ever-lasting transformation as the conditions change.

No description available.

Syftet med den här uppsatsen är att undersöka skillnader i tolkningen av karaktärer i två versioner av en och samma scen klippta utifrån olika metoder. Den ena klipptes utifrån kontinuitetsprincipen och den andra klipptes inte alls, utan gestaltades istället enbart genom en helbild, i en tagning, i vad som kallas en sequence shot. Genom att gestalta ett maktspel mellan tre karaktärer i scenen tar författarna fasta på skillnader i uppfattningen av makt- och statusförhållanden. Versionerna sågs av två olika testgrupper i omvänd ordning som sedan fick frågor i enkätform. Resultatet visar på att kontinuitetsklippningen ger tittaren en mer enhetlig bild av handlingen och fördjupade bilder av karaktärerna, medan en sequence shot ger en större spridning tolkningar och en mer övergripande bild av karaktärernas gruppdynamik i scenen.

Cutting tool inserts are often coated with thin layers either through chemical vapor deposition (CVD) or physical vapor deposition (PVD) processes. In order to have a better wear resistance cutting tools are mostly subjected to post-coating treatment processes. Shot peening is one of the processes that is used to improve the fatigue life of metallic components. In this study, the finite element (FE) method is employed to model the elastic-plastic deformation and development of residual stress distributions in a cutting tool after the impact of a shot medium. To carry out the work, CVD coated cemented carbide has been chosen to be the workpiece (insert), and the coatings of the chosen insert are Titanium Carbo Nitride (TiCN)  and Aluminum Oxide (Al2O3). Aim of the study is to model a single impact in the shot peening process on a surface of a coated cemented carbide insert while simulating the plastic deformation of the materials. In addition, the objective of the study is also to understand and explain the mechanics of shot peening process and find applicable mechanical properties of the materials for FE modeling. Conjugately, the influence of shot peening process parameters (e.g. velocity, diameter or shape of the peening media) on residual stress distribution has been investigated and the results obtained were compared to the one observed from experiment. The modeling in the study is carried out both with and without initial residual stresses in the materials. The initial residual stresses are estimated by applying a thermal load to the model. The results show that the compressive residual stresses achieved while shot peening by an edge-shaped medium are significantly higher at the surface (i.e. in a coated layer) than compared to a globular medium. In contrast, it is observed that the compressive residual stresses in the cemented carbide are significantly higher and deeper when shot peened with globular medium than the edge-shaped medium. Furthermore, the results of parameteric study demonstrate that the smaller medium induces higher residual stresses at the surface (i.e. in a coated layer) than in the cemented carbide. In contrast, it is observed that the bigger medium induces less residual stresses at the surface (i.e. in a coated layer) and higher residual stresses deeper in the cemented carbide. Whereas, it is observed that the higher residual stresses at the surface (i.e. in a coated layer) and in the cemeneted carbide can be achieved simultaneously by shot peens having a  higher velocity. Residual stress profiles modelled in this report correlate with data from previous studies.   This study has been carried out at Sandvik Coromant, Edge and Surfaces department in Stockholm, Sweden.

Title: Analysis of the badminton matches in men's and women's singles The goals of essay: Main goal of the essay is to find out differences in first three shots and its most common combinations in single matches of men and women in relation to gaining a point during the rally. Another goal was to find frequency of using backhand and forehand hold in second and third shot. The last goal was to compare the frequency of ending in first, second and third shot between men and women. Methods: Indirect analysis - videoanalysis Results: According to results men and women primarily choose different types of serving. Men most often use short backhand serve, women high forehand serve. For the second shot men choose short shot or lob, women use smash. The third shot is lob in men rally, women choose short shot most often. Generally men uses in the first three shots shots less clears than women do. For both sexes are the most effective shots offensive ones, but men are more successful than women. The most effective direction of return was for men and for women the body of the player, in third shot it was the backhand side too (left middle and left back of the field). Men were more successful with forenhand, women with backhand. The frequenc of endings to third shots shows that women more often end the rally...

Title: Comparison of the badminton match in men's and women's singles Objectives: The goal is identify the most effective a choice combinations of the first three shots, leading to the acquisition point in singles between the best Czech male and female players. And also to discover determine and compare which was the usage of different types of serve, return and the third shots. Methods: Comparison of differences in badminton match between genders was based on indirect observation - video analysis. Statistics were used for data processing. Results: The results shower huge difference in tactical approaches among men and women players. Women preferred long high service and most of the service returns were played into back backhand court area and the third shot usually routed into back forehand court area. It means that the most played shot was a clear shot and the most occupied part of court was a back backhand corner. Despite more successful was when women players played drop by their third shot. Domination of short backhand service was seen by men players, the second shot routed the same direction like women players second shot - to the back backhand court, but also to the front part of the court. Third shot was mostly played into back backhand part of the court. If players wanted to be successful,...

abstract: High-level inference tasks in video applications such as recognition, video retrieval, and zero-shot classification have become an active research area in recent years. One fundamental requirement for such applications is to extract high-quality features that maintain high-level information in the videos. Many video feature extraction algorithms have been purposed, such as STIP, HOG3D, and Dense Trajectories. These algorithms are often referred to as “handcrafted” features as they were deliberately designed based on some reasonable considerations. However, these algorithms may fail when dealing with high-level tasks or complex scene videos. Due to the success of using deep convolution neural networks (CNNs) to extract global representations for static images, researchers have been using similar techniques to tackle video contents. Typical techniques first extract spatial features by processing raw images using deep convolution architectures designed for static image classifications. Then simple average, concatenation or classifier-based fusion/pooling methods are applied to the extracted features. I argue that features extracted in such ways do not acquire enough representative information since videos, unlike images, should be characterized as a temporal sequence of semantically coherent visual contents and thus need to be represented in a manner considering both semantic and spatio-temporal information. In this thesis, I propose a novel architecture to learn semantic spatio-temporal embedding for videos to support high-level video analysis. The proposed method encodes video spatial and temporal information separately by employing a deep architecture consisting of two channels of convolutional neural networks (capturing appearance and local motion) followed by their corresponding Fully Connected Gated Recurrent Unit (FC-GRU) encoders for capturing longer-term temporal structure of the CNN features. The resultant spatio-temporal representation (a vector) is used to learn a mapping via a Fully Connected Multilayer Perceptron (FC-MLP) to the word2vec semantic embedding space, leading to a semantic interpretation of the video vector that supports high-level analysis. I evaluate the usefulness and effectiveness of this new video representation by conducting experiments on action recognition, zero-shot video classification, and semantic video retrieval (word-to-video) retrieval, using the UCF101 action recognition dataset. / Dissertation/Thesis / Masters Thesis Computer Science 2016